Critics Question if Humans Keep Control as AI Speeds up Military Targeting
Witnesses told US lawmakers that artificial intelligence may compress military decision-making timelines so significantly that human oversight becomes nominal rather than functional.
Witnesses testified before the Tom Lantos Human Rights Commission that the rapid integration of artificial intelligence into military operations may outpace the human capacity for meaningful verification. According to the testimony, while current regulations require human approval for autonomous weapons, the speed at which AI systems process data and recommend targets could render that approval procedural rather than substantive. The core concern raised is not the existence of fully autonomous weapons, but the subtle influence AI exerts on human decision-makers by drastically reducing the time available for contextual analysis and ethical deliberation.
The hearing, focused on the implications of AI for human rights, highlighted a growing tension between operational efficiency and accountability. Anna Mysyshyn, a Ukrainian AI-governance researcher, argued that the mere presence of a human in the loop does not guarantee effective control. She noted that in conflict zones like Ukraine, AI systems have already begun compressing decision cycles from hours to minutes, creating scenarios where operators lack the bandwidth to challenge incorrect recommendations before harm occurs. This shift challenges the traditional understanding of human oversight in warfare.
Pentagon Strategy Prioritizes Speed
The Department of Defense has explicitly embraced these technological shifts in its recent strategic documents. The Pentagon’s 2023 AI adoption strategy identified the creation of "fast, precise and resilient kill chains" as a primary objective, signaling a desire to accelerate every stage of military engagement. Furthermore, the 2026 AI strategy outlines a vision for an "AI-first" warfighting force, aiming to redesign operational workflows around current machine learning capabilities. According to GN geopolitics/rights (en-US), this institutional push toward automation contrasts sharply with the ethical concerns raised by civil society organizations during the congressional hearing.
Critics argue that this strategic direction risks normalizing a level of automation that bypasses necessary human discernment. Amanda Klasing, national director of government relations at Amnesty International USA, warned that AI can compress the military kill chain to a degree where it becomes difficult to determine if genuine human judgment was exercised before a strike. The concern is that the speed of algorithmic recommendation may psychologically override human hesitation, leading to decisions that are technically approved by a person but effectively dictated by the system.
Limits Of Statistical Inference
Beyond the speed of decision-making, experts have raised concerns about the reliability and auditability of AI systems. Joseph Chapa, a former military officer and scholar of military ethics, urged policymakers to look beyond simple autonomy levels. He argued that statistical AI systems can produce substantially different outputs from relatively small changes in inputs, a behavior distinct from traditional software. This unpredictability makes it difficult to reconstruct how a specific recommendation was generated after the fact, complicating efforts to audit system performance or identify errors in operational environments.
Chapa emphasized that the current regulatory framework, which mandates human judgment over the use of force, may not account for these specific technical vulnerabilities. He called for expanded AI education within the military, particularly for personnel involved in targeting and intelligence. Without a deeper understanding of how these systems function and fail, human operators may be unable to recognize when a recommendation is flawed, effectively stripping them of the ability to exercise meaningful control.
Accountability Gaps In Civilian Harm
The integration of AI also exacerbates existing challenges in mitigating civilian harm. The Defense Department established a Civilian Harm Mitigation and Response Action Plan in 2022, including a dedicated Center of Excellence for analysis and training. However, a review by the Defense Department inspector general in May identified significant implementation failures. According to the report, the department faced issues such as the loss or reassignment of key personnel, the termination of funding for a civilian-harm data-management platform, and the cessation of meetings by a senior steering committee.
These administrative lapses create a critical gap in accountability as AI systems become more prevalent. Klasing noted that the military currently has limited ability to track and report civilian harm, a deficit that becomes more dangerous when decision-making speeds increase. The forward question for policymakers is whether the existing legal and institutional frameworks can adapt to a future where AI not only influences targeting but fundamentally alters the timeline and transparency of military operations. Observers will be watching for legislative responses that address both the technical auditability of AI systems and the structural integrity of civilian protection mechanisms.
